Configuration reference Everything lives in .personaqa/config.yaml. Only version and app are required; every other section has defaults. /personaqa:init writes a commented starting point, and /personaqa:validate checks the file against the schema .
Any string value can use ${ENV_VAR} or ${ENV_VAR:-default}.
app Key Default Notes name required Shown in reports base_url required Where mission start_url paths are resolved. --base-url on a run overrides it allowed_domains required Hosts the browser may load, as host, host:port or *.example.com. Everything else is refused, including links, images and scripts. Add payment or sign-in providers your journeys use
auth See Logins .
Key Default Notes strategy none none, storage_state or login_mission storage_state none Saved login file, for example .personaqa/auth/default.json login_mission none Mission id that signs in accounts.<key> none email, password and optional storage_state
runner Key Default Notes max_parallel 3 Personas running at once (1 to 10) max_steps_per_run 60 Actions before a session ends as limit. A mission’s max_steps overrides it max_minutes_per_run 8 Minutes before a session ends as limit. A mission’s time_limit_minutes overrides it model sonnet Model for persona runners, for example haiku to cut cost. See Cost and usage
personas and missions Key Default Notes personas.enabled all all or a list of ids run when no personas are named personas.paths [.personaqa/personas] Folders searched before the built-in cast missions.paths [.personaqa/missions] Folders holding missions
output Key Default Notes dir .personaqa/runs Where run folders go (gitignored by /personaqa:init) screenshots on_issue every_action, on_issue (issues, accessibility problems and the final screen) or off trace on_failure Playwright trace: always, on_failure (sessions that did not succeed) or off regression_tests suggest Reserved; not used yet. Write tests on request with /personaqa:test
issues Key Default Notes sink file github makes /personaqa:report file (and the nightly workflow) file every finding at or above min_severity, not only confirmed ones github.labels [personaqa] Added to every issue, alongside a11y or ux and wcag-x.y.z labels github.min_severity medium low, medium, high or critical
ci Named sets for /personaqa:run --set <name>:
ci :
pull_request :
personas : [ james , priya ]
missions : [ checkout ]
nightly :
personas : all
missions : all
inspectors Key Default Notes a11y.enabled true Accessibility inspector, and axe-core scans during runs a11y.standard wcag22 wcag21 or wcag22 a11y.level AA A, AA or AAA a11y.mode ride_along ride_along scans every new page a persona reaches; sweep only scans in /personaqa:audit; both a11y.ignore [] axe rule ids to skip, for example color-contrast-enhanced ux.enabled true UX inspector ux.heuristics nielsen Nielsen’s ten usability heuristics
Turn inspectors off for one run with --no-inspectors, or run only one with --with a11y.